Some Remarks by George Hart on Herman Tieken's Kavya in South India

Some Remarks on
Herman Tieken's Kavya in South India

George Hart, Professor of Tamil and Chair of Tamil Studies
University of California, Berkeley, USA 2001

Recently, a book by Herman Tieken has appeared entitled Kavya in South India: Old Tamil Cankam Poetry, published by Egbert Forsten in Groningen, The Netherlands, 2001. The book is part of the prestigious Gonda Indological Studies, and so must be taken extremely seriously. I am currently reviewing the book for the Journal of the American Oriental Society, but I feel it is necessary to give some information about it here, as the book claims to overturn virtually all of our ideas about premodern Tamil literature.

Here is some information on the book, taken from the Indology forum:

Did Tamil Cankam poetry describe a contemporary society, or an idealized pure Tamil society of the past, as it was imagined in a time already greatly influenced by North-Indian Sanskrit culture ... ?

The following recent publication is perhaps of interest to readers of this list because of the challenging thesis on the relative chronology of early Tamil poetry and Sanskrit Kaavya defended in it.

Title: Kaavya in South India: Old Tamil Cankam Poetry. Author: Herman Tieken Publ.: Groningen: Egbert Forsten, 2001

From the back cover:

"Old Tamil Cankam poetry consists of eight anthologies of short poems on love and war, and a treatise on grammar and poetics. The main part of this corpus has generally been dated to the first centuries AD and is believed to be the product of a native Tamil culture.

The present study argues that the poems do not describe a contemporary society but a society from the past or one not yet affected by North-Indian Sanskrit culture. Consequently the main argument for the current early dating of Cankam poetry is no longer valid. Furthermore, on the basis of a study of the historical setting of the heroic poems and of the role of Tamil as a literary language in the Cankam corpus, it is argued that the poetic tradition was developed by the PaaNTiyas in the ninth or tenth century. ... ... the identification of the various genres of Cankam poetry with literary types from the Sanskrit Kaavya tradition ... indicates that in Cankam poetry Tamil has been specifically assigned the role of a Praakrit. ... "

My edition has a slightly different blurb on the back, but it is not significantly different. Here is what I replied to this notice before I had seen the book:

I notice the message reproduced at the end of these comments in Indology in July. While I have not yet seen Prof. Tieken's book, I can certainly respond to the ideas on the back cover. There is overwhelming and indisputable evidence that the anthologies were not written as late as the ninth or tenth century. It is, moreover, quite certain that the Sangam anthologies were not patterned after Prakrit or Sanskrit. Let me make a few brief points.

1. Language. This is absolutely indisputable. The Sangam texts use very different (and demonstrably much more archaic) forms and vocabulary than later texts of the sixth century onward. Many verb forms that disappeared by the sixth century can be shown to be part of an older Dravidian verb system that is quite consistent with an age of the second or third century. The vocabulary is also quite different from that of later works -- it has much less Sanskrit, and uses words that are not attested later. A good source for these older forms is V. S. Rajam's A Reference Grammar of Classical Tamil Poetry : 150 B.C.-pre-fifth/sixth Century A.D. Philadelphia, Pa., American Philosophical Society, 1992.2. Content. While the akam (love) poems share some themes with Prakrit and even Sanskrit (as I have shown), they are still radically unlike the poems in those languages. Prakrit has nothing like the five tiNais, and it is not nearly as carefully worked out, with stock speakers, stock images for each landscape, ragas (called paNs) for each, and the like. There is absolutely nothing like the PuRam (heroic) poems of the PuRananuru and the Patirruppattu in Prakrit or Sanskrit. That is because the Puram poems are mostly written as imitations of the productions of low-caste bards and drummers.

3. Culture. The poems show a coherent culture that is utterly different from the 9th or 10th century. It is clear, if one reads the Purananuru, that the poems are directly about events the authors have heard of. Many of the poems concern marginal people at the borders of society. This is not the case of the Sanskrit or Prakrit traditions. Where is there anything like the famous Kalittokai poem of the lame man pursuing the hunchback woman?

4. History. The poems name hundreds of poets and kings -- and string them together in a narrative that is chronologically coherent. The names are quite unlike the names of the 9th and 10th century. There are many historical facts that have been confirmed by archeological and other evidence -- some kings who appear on coins, or in datable contemporary inscriptions (1st-3rd century), Roman coins, description of trade also found in outside sources, and the like.

5. Literary theory and usage. The Tolkappiyam describes theories and systems that are mostly quite foreign to Sanskrit and Prakrit, but which fit Sangam literature quite well. Its grammar describes some forms that are quite old (as shown by the earliest inscriptions), and even the writing system it describes, with the puLLi, has now been shown to be as early as the second century or even earlier. The Sangam poems do not use anything related to Sanskrit meter, unlike the poems of later times. By the ninth and tenth centuries, almost all literature in Tamil divided its stanzas into four parts, like Sanskrit and Prakrit (though they never actually borrowed Sanskrit meters).

6. Hindu, Buddhist, and Jaina elements. In the Sangam poems, Murugan has not yet been identified with Kartikeya -- he is a folk spirit that possesses people and must be propitiated. There is not much mention of Visnu or Siva, while it is clear that Jainism and Buddhism are both present in Tamil Nadu. Many of the gods are local and do not appear in later literature. All of this accords perfectly with what we know of that period, and does not fit at all the later period.

This is only a cursory response -- it seems almost a waste of time to go on, as the evidence is so abundant and convincing. The fact is, the poems are quite unaware of Prakrit or Sanskrit literature -- though they do know of the Ramayana and the Mahabharata (which fits with their accepted date of 1st-3rd century AD). They do not resemble Prakrit or Sanskrit literature enough to be modeled upon them -- I have argued that both Tamil and Prakrit/Sanskrit use the same body of conventions, which they got through southern (Tamil and Maharastri) folk traditions. But the poems themselves are quite different and work in very different ways. Finally, there is a convincing -- and enormous -- body of coherent and mutually reinforcing historical, linguistic, cultural, religious and literary evidence that shows clearly the poems are much earlier than the 9th or 10th century.

Yes, some 9th or 10th century poets might have decided to write some "old" literature based on Prakrit and Sanskrit. But would they have invented hundreds of archaic forms and words that fit the development of Dravidian? Would they have eschewed Prakrit/Sanskrit ideas and metrical patterns? Would they have carefully gotten rid of almost all their Sanskrit words and invented hundreds of words that are not found in the other 9th century literature? Would they have made up the names of hundreds of poems and kings and woven them into a huge corpus that is chronologically consistent (and fits with inscriptional and numismatic evidence)? Tieken's argument (if it is correctly reflected in the blurb on the back cover) just does not make sense.

It is as if one were to claim the Vedas were written in the 10th century AD by a group of people who wanted to reflect an idealized past. Indeed, the Sangam works contain much more historical information than the Vedas -- it would be much easier to 'prove' that the Vedas were written in the 10th century than that the Sangam poems were.

What is it about some European Sanskritists that makes them unwilling to accept that a non Indo-European people could create a great literature on their own in South Asia? The evidence of the non-derivative nature of Sangam literature is absolutely convincing. I hope that some will read the translation of the Purananuru that Heifetz and I recently published. How these poems could be derived from the Sanskrit/Prakrit tradition utterly mystifies me -- and I have read most of the kavya literature (Kalidasa, Magha, Bharavi) in the original. And I have read the Prakrit poems with the chaayaa anuvaada. By the way, the Purananuru is one of the seminal texts of premodern India -- it is quite as important as the epics and the Vedas for understanding the development of South Asian culture.

Here are some further comments after looking over his book. One of Tieken's arguments -- in fact, his most important argument -- is statistical: he claims that the arrangement of the poems is not random (as they presumably would be if anthologized), but rather contains repeated words and phrases between contiguous poems (9 poems forward or backward), and that means the poems must have been all composed in order by one person who unconsciously -- or consciously -- reused phrases from what he had just composed. As evidence for this, he lists repeated words/phrases that recur between neighboring poems.

He writes, "Apart from the difficulties in finding a poem meeting these requirements all at the same time, we should consider the size of the corpus which the compiler was supposed to have had at his disposal. If we set the frequency of a word, rather arbitrarily, at once every tenth poem, the compiler would in theory have required ten poems to hit upon another occurrence of aNNal, ten times ten poems in order to find one which in addition contained eeRu, a thousand poems to find one which also contained punkaNmaalai, ten times thousand (sic) poems to find one which also contained peyar, and so on. It is hardly unreasonable to doubt if the compiler was ever interested in memorizing such a vast corpus of poems...... " The fact is, this is an elementary error in statistics.

Here is the proper statistical calculation: Suppose the anthologies contain a total of 8000 different words and 80 words per poem. The likelihood of word 1 in poem 1 being word 1 in poem 2 is 1/8000. The likelihood of its being any word in poem 2 is 80/8000=1/100. Likelihood of 1.2 being any word in poem 2 is also 1/100. Thus the chance of any word in poem 1 being any word in poem 2 is 80/100=80%. If you include 2 poems, chance of a word in poem 1 being any word in poem 2 or 3 is 80/100+80/100=160%. If you include 8 poems forward and backwards, chance is 80*16=1280%. That is to say, you would have to go through 13 poems before you found one that did not have any words in common. The chance of 2 words occurring are 1280/2, of 3 words are 1280/4. In fact, it is most likely that 13 words will be found to exist within the preceding and following 8 poems. (In fact, we might expect even more repetition, as the poems are based on oral models and use conventional phrases and themes).

In his work, Tieken gives a 10 page appendix in which he shows that words repeat between contiguous poems (including 9 forward and 9 backward). However, he does not do what any standard scientific study would do: he does not compare his results with the results of poems chosen at random. I did so, choosing poems from the Kuruntokai at random. Let me first give Tieken's correspondences for poems 140-149:

140 cel, koLa, maakkaL, azintu, aRintu, puu
141 celku kooL kiLi kaTiiiyar kaT naaL kai naaTa paTu
142 puu kiLi kaTiyum kaT paanaaT aRintanaLoo
143 uTaiyan uriyatu malai icai ankaluz naaTan paaTTaaLan
144 cenranaL uriyatu malai nanree veNTalai pirivil naaTTee paTutalum
145 uTaittee tuyiR maakkaL kaN paanaaL aanaa ciRu
146 pirintoorppuNarpoor maakkaL nanRu veNTalai uTai taNTu
147 pirintooree tuyil kanavee anna ankaluz mayir
148 kanavoo anna
149 oonku aLitoo niiTu tara kai ciRu puun nillaavee

Note that this actually should give more correspondences than mine, as it is based on using the next ten poems as well and some of the correspondences (e.g. oonku in 149) refer to later poems. Here is the result I obtained by using ten random KuRuntokai poems (referring only to one another):

5 kaN kol toozi tirai kaama nooy utai tirai tivalai niir pulampan ena pal uNkaN paatu mellam
186 kaN kollai toozi koTi mullai talai ena aar talai mel punatta mullai koti naataRku
240 koTi tirai mullai talai nooy tirai pal talai toozi mullai koti pani kiLi verukku vantatan pora toonRi maalai avar maNi netu oLi
253 toozi nivanta keeLaar aayinum |puu ceer| peru kavin nin tuyar |oli kazai| malai pukaa kal cel iRantoor ooíku
308 talai kaama utai naatan tuyar malai iru piti izi cilampin kaN |maa malai| kai
367 koTiyoor toozi kotiyoor vanticin toonRum avar maNiyin aayinum puucal kavin nin malai pukanRu kal iRaiya izi |maa malai| toti peRiiiyar |uva kaaN| naattu |taN naRu| ooíkiya nalkaar
390 ena paatu vantu netu keeLaay cellaatiim piti taNNumai ciRu peyarum munai
133 pulampinan uNta (see uN in uN kaN) toozi punavan kiLi peru olittu iruvi ciRu peyal olittu nalam
7 kol mel pora olikkum cilampee totiyooL olikkum kaalana nalloor munniyoor
398 tivalai ena uNkaN toozi pani vantena maalai tuyar kaN kai peRal uvakaiyin taNNena peytu kaalai aaka
107 niir verukiRku toonRi netu iruL oL aaki

As you can see,I actually found many more correspondences in non-contiguous poems than Tieken found in contiguous ones. The central argument of his book-that there are more repetitions between contiguous poems than non-contiguous ones-is thus shown not to be valid. There is much more to be said about this book. Suffice it to say at present that I have found nothing whatsoever that would lead me to accept any of its premises.